JP3630725B2 - Three-dimensional shape input device and input method thereof - Google Patents

Description

[0001]
[Industrial application fields]
The present invention relates to a three-dimensional shape input device for obtaining a three-dimensional geometric shape of an object in an image from a stereo image and an input method thereof.
[0002]
[Prior art]
As a system for obtaining a three-dimensional shape of an object in an image from a stereo image taken by a stereo camera or the like, a method has been proposed in which a human obtains a three-dimensional shape while interactively specifying corresponding points (Japanese Patent Laid-Open No. Hei. No. 5-299515). In this method, 3D coordinates are generated from corresponding points on a stereo image based on the principle of triangulation, and then a 3D surface is generated by sequentially generating a network of triangles connecting these 3D coordinates. It is to do. As a method for generating a triangle, a triangular network generating method for a stereo image based on the Delaunay principle is generally used. The specified corresponding point on the two-dimensional plane is covered with the Delaunay triangle network, and a three-dimensional plane is constructed from the connection information of this triangle network and the three-dimensional coordinates of the corresponding points calculated by the principle of triangulation. It is to do.
[0003]
In addition to the above proposals, it has also been proposed that textures attached to the surface of a three-dimensional shape are taken from a stereo image during editing.
[0004]
The three-dimensional shape generation method is effective for sequential point addition, and the triangle network is recalculated and updated each time a point is added. The advantage of this method is that humans can proceed with editing while confirming stereo images and 3D shapes.
[0005]
[Problems to be solved by the invention]
However, the method of interactively designating corresponding points on a stereo image has the following problems. That is, when the number of corresponding points (becomes the vertices of a triangle) increases, the relationship between the corresponding points tends to be lost. That is, all the vertices have corresponding points displayed on both the left and right stereo images and the three-dimensional display, but in the case where the vertices are complicated, a certain vertex is displayed on these three types of displays. It is difficult to tell which point corresponds to each. In particular, when the points are dense, it is easy to mistake the correspondence of the designated points.
[0006]
In addition, an error often occurs in the three-dimensional shape due to lens distortion at the time of capturing a stereo image, a specified position shift of a corresponding point on the stereo image, or the like. In such a case, currently, the corresponding point is deleted once and re-applied. However, if the correspondence is deleted, the connection relationship of the triangle network may change, and the three-dimensional shape obtained by retyping is difficult to predict. Furthermore, in stereo images, small lateral shifts of corresponding points often have a large effect on the three-dimensional shape. For this reason, it has been difficult to intuitively understand the correspondence between the operation and the three-dimensional shape.
[0007]
Therefore, the present invention has dense vertices. Also It is an object of the present invention to provide a three-dimensional shape input device and an input method thereof that make it easy to understand the relationship between the corresponding points and easily change the position of the corresponding points.
[0008]
[Means for Solving the Problems]
In order to achieve the above object, a three-dimensional shape input apparatus according to claim 1 of the present invention comprises an image display means for displaying a stereo image, a three-dimensional shape data display means for displaying three-dimensional shape data, and the stereo. Point designation means for designating corresponding points in the left and right images of the image, calculation means for sequentially calculating three-dimensional coordinates from the corresponding points and the camera parameters of the stereo image, and coordinates on the input image Generating means for generating a triangular network by connecting a set of three points, and surface forming means for approximately forming the surface of the object using the generated triangular network and the calculated three-dimensional coordinates Superimposing means for superimposing the generated triangular mesh on the stereo image and the three-dimensional shape data, and specifying the vertex of the triangular mesh superimposed on the image display means and the three-dimensional shape data displaying means And vertex designation means for performing, By the vertex designation means According to the specification On the display of the stereo image and the three-dimensional shape data, And highlighting means for highlighting corresponding vertices.
[0009]
The three-dimensional shape input apparatus according to claim 2 is the three-dimensional shape input apparatus according to claim 1, wherein the vertex position on each stereo image is Move Moving means to move, Said The calculation means calculates the coordinates of the vertices of the triangle network displayed superimposed on the image display means and the three-dimensional shape data display means while corresponding to each other as the vertices move.
[0010]
The three-dimensional input method according to claim 3 displays a stereo image, displays three-dimensional shape data, designates corresponding points in the left and right images of the stereo image, and sets the corresponding points and camera parameters of the stereo image. 3D coordinates are sequentially calculated from the above, and a triangle network is generated by connecting the coordinates on the input image in pairs, and the generated triangle network and the calculated 3D coordinates are generated. Are used to approximate the surface of the object, and the generated triangle network is superimposed on the stereo image and the three-dimensional shape data, and the triangle is displayed superimposed on the stereo image and the three-dimensional shape data. Specify the vertex of the net, and according to the specification On the display of the stereo image and the three-dimensional shape data, Highlight the corresponding vertex.
[0011]
[Action]
In the three-dimensional shape input device according to claim 1 of the present invention, the stereo image is displayed by the image display means, the three-dimensional shape data is displayed by the three-dimensional shape data display means, and the correspondence in the left and right images of the stereo image is displayed. A point is designated by point designating means, three-dimensional coordinates are sequentially calculated by the calculating means from the corresponding points and the camera parameters of the stereo image, and the coordinates on the image input by the generating means are made into a set of three points. A surface of the object is approximately formed by the surface forming means using the generated triangular network and the calculated three-dimensional coordinates, and the generated triangular network is generated. Is superimposed on the stereo image and the three-dimensional shape data by the superimposing means, and the top of the triangular mesh displayed on the image display means and the three-dimensional shape data display means by the vertex specifying means. Performs a specified, by highlighting means By the vertex designation means According to the specification On the display of the stereo image and the three-dimensional shape data, Highlight the corresponding vertex.
[0012]
【Example】
Next, an embodiment of the three-dimensional shape input device and the input method of the present invention will be described.
[0013]
FIG. 2 is a block diagram illustrating a configuration of the three-dimensional shape input apparatus according to the embodiment. The three-dimensional shape input device of this embodiment includes a CPU 14, a main storage device 13, a hard disk device 12 that records a stereo image 21, a bitmap display 15, a keyboard 16, a mouse 17, and an input / output unit 11. In the three-dimensional shape input device, stereo images taken by two cameras (not shown) are read into the hard disk device 12 through the input / output unit 11 when digitized.
[0014]
FIG. 1 is an explanatory diagram showing a screen on which a stereo image is displayed. In the figure, 3 and 4 are windows for displaying left and right images constituting a stereo image, respectively. Reference numeral 5 denotes a window for three-dimensionally displaying the three-dimensional shape data obtained by the interactive processing.
[0015]
In general, a stereo image indicates two images taken at an angle close to human parallax, but in this embodiment, it is understood in a broader sense and used to mean two images taken from an arbitrary angle. To do.
[0016]
4 and 5 are flowcharts showing a three-dimensional shape input processing routine. FIG. 3 is an explanatory diagram showing a data structure used in the three-dimensional shape input processing routine. First, when the three-dimensional shape input processing device is activated, the stereo image 21 is read from the hard disk device 12 to the main memory 13 (step S402), and the read left and right images are displayed on the bitmap display 15 as indicated by 3 and 4 in FIG. Are displayed in a multi-window (step S403).
[0017]
Next, the operation waits for an operation such as mouse input, keyboard input, menu selection, etc. from the operator (step S404). If there is an operation, processing corresponding to the operation is executed, and then the operation waits again in step S404. When the operation is mode selection, the mode is switched by keyboard input, menu selection, or the like (steps S406 and S408). If the operation is a mouse click on the window images 3, 4, and 5 shown in FIG. 1, appropriate processing is executed according to the mode (steps S409 to S411). If the operation is neither a mouse click nor a mode selection, another process corresponding to the operation is performed (step S407), and the process returns to step S404. Hereinafter, the processing for each mode will be described.
[0018]
[Coordinate specification (add point) processing with mouse]
When the mode is the addition mode in step S409, the position of the mouse when the mouse button is clicked on the stereo image displayed on the bitmap display 15 is acquired (step S413). It is confirmed whether or not coordinate designation has been performed on both the right window image 4 and the left window image 3 (step S414). The operator designates the coordinates of the same point on the object appearing in the left and right images with the mouse. Here, the designated point becomes a vertex of the three-dimensional shape by the subsequent three-dimensional coordinate calculation processing.
[0019]
In this embodiment, the left and right corresponding points are both manually specified by visual inspection. However, if one image point is specified by pattern matching or the like, the other corresponding point is automatically calculated and obtained. It is also possible to do so.
[0020]
When the coordinates on the left and right images are aligned, the coordinates are stored (step S415). If not, the process returns to step S413 and the same process is performed. The left and right designated coordinates are stored as a data structure shown in FIG. That is, a list (A) (hereinafter referred to as a point list) having three-dimensional coordinates obtained by calculation described later from the designated coordinates on the left image, the designated coordinates on the right image, and the left and right designated coordinates. register. Thereafter, when the left and right points are designated, the coordinates are sequentially added and registered in this list. The i-th list element in FIG. 3 includes three coordinates corresponding to the same point added to the i-th, that is, designated coordinates (Lxi, Lyi) on the left image, and designated coordinates (Rxi) on the right image. , Ryi), three-dimensional coordinates (Vxi, Vyi, Vzi) obtained by calculation described later from the left and right designated coordinates.
[0021]
Further, the processing of steps S417 to S422 is performed to perform three-dimensional coordinate calculation, surface generation, and three-dimensional display. Here, the i-th left and right corresponding points are assumed to be given. In step S417, three-dimensional coordinates (Vxi, Vyi, Vzi) are calculated using the triangulation principle from the left and right coordinates (Lxi, Lyi), (Rxi, Ryi) of the i-th point list and predetermined camera parameters. Then, it is registered in the i-th point list (step S417). The camera parameter is a parameter that includes stereo image shooting conditions, that is, information including camera position, direction, focal length, etc., for each left and right image (for each camera that shoots left and right images). Have it separately. However, the principle of triangulation can be applied when parallel viewing is assumed, but even if parallel viewing is not assumed, the camera direction, three-dimensional position, and focus when shooting from any two directions If lens information such as distance is given as camera parameters and corresponding points on the left and right stereo images, the three-dimensional coordinates of the points can be calculated in principle. This camera parameter is measured in advance and given to the system by some method such as reading from a file.
[0022]
Next, a surface is generated from the point list of FIG. 3A (step S418). In this embodiment, a triangular patch is generated from a specified point sequence on one of the images, and is treated as a surface. That is, for example, three points are selected from the sequence of points (Rx1, Ry1), (Rx2, Ry2),..., (Rxn, Ryn) on the right image (two-dimensional plane) so that the connections do not overlap. A triangular patch is generated. For all the input points, one triangle is generated when the first three points are specified, and after the fourth point, one point added every time one point is added and one point already input These two points are selected so that the triangles do not overlap, and a new triangle is generated. Re-evaluate triangles that have already been generated as needed when points are added, and regenerate all (or some) triangles so that they do not overlap.
[0023]
Here, as a method of generating a triangle, a method of automatically generating a triangle network based on the Delaunay principle is used. Triangular network generation based on the Delaunay principle is to generate a triangular network by connecting point groups so that the smallest vertex angle of a triangle is maximized, that is, a triangle close to a regular triangle is formed. This method is very effective as a method for automatically generating a triangular network from points such as the amount of calculation for re-evaluation accompanying the addition of points. However, since all triangle networks are automatically generated, it is not possible to leave an edge connecting two points explicitly. A method for solving this problem is disclosed in Japanese Patent Laid-Open No. 5-299515.
[0024]
FIG. 1 is an example of a screen showing a state where a triangular patch is pasted on an image, and 4 shows a state of a triangular patch in the right image. The plane on the left image and in the three-dimensional space can be obtained as follows. For example, when attention is paid to one triangle connecting the points (Rx1, Ry1), (Rx2, Ry2), (Rx3, Ry3), on the corresponding left image coordinates from the list elements P1, P2, P3 of the corresponding point list and The point of the three-dimensional coordinate is taken out, and the triangle on the left image and in the three-dimensional space is obtained.
[0025]
The triangle generated in this manner is added to the surface list having the structure of FIG. 3B (step S419). Each element of the surface list includes pointers to the three elements of the point list (FIG. 3A) corresponding to the coordinates of the three vertices of the triangle, and is included in the surface list and the surface list element. It is possible to generate two-dimensional and three-dimensional triangular patches by following the point pointer. This face list constitutes the Delaunay triangle network.
[0026]
The two-dimensional triangular patch is displayed so as to be superimposed on the images indicated by 3 and 4 in FIG. 1 (step S420).
[0027]
The three-dimensional triangular patch is displayed three-dimensionally as indicated by 5 in FIG. As a method of displaying a three-dimensional object, several methods such as wire frame display and shading display are known. However, the three-dimensional display method is not particularly particular. By repeating the above operation, the three-dimensional shape of the object surface can be obtained.
[0028]
[Vertex highlight processing]
The vertex emphasis display processing is a triangle on any one of the three display windows from right to left stereo image display (4, 3 in FIG. 1) to three-dimensional display (5 in FIG. 1). When the vertex of a patch is designated with the mouse, the corresponding vertex is highlighted on other displays.
[0029]
If the mode is the search mode in step S410, it is checked whether there is an already highlighted vertex (step S422). Whether or not there is a highlighted vertex is determined by looking at the contents of a buffer (hereinafter referred to as a designated point buffer) prepared for storing the designated point. If there is a highlighted vertex, the vertex is reset to normal vertex display (step S423), and the designated point buffer is cleared (step S424). Next, the position of the mouse click is acquired (step S425), and the nearest vertex is designated (step S426). When the mouse click is the window image 3 or 4, the distance between the coordinates of the click position and all the vertices may be compared. Even in the case of a three-dimensional display window, for example, the designated point may be determined by comparing the distances using the vertex positions of the screen display projected in two dimensions.
[0030]
The designated point is searched from the point list (step S427), and it is determined whether or not the designated point is found (step S428). If the designated point is not found, the process returns to step S404 and the same processing is performed. If the designated point is found, the vertex information on the left and right images and the vertex on the three-dimensional display corresponding to the designated point are switched to highlighted using the coordinate information of the point list (step S429), and the designated point buffer stores the specified point. The point (list address) is stored (step S430). Note that the highlighting is displayed so that the appearance is different from the normal vertex by changing the color or displaying the brightness.
[0031]
[Deleting vertex]
If it is determined in step S411 that the deletion mode is set, it is determined whether there is a highlighted vertex (step S431). If there is a highlighted vertex, this point is removed from the point list as a deletion target (step S432), and the face list is regenerated (step S434) by regenerating the triangle network (step S433). ), The display is updated (step S435).
[0032]
As described above, according to the three-dimensional shape input device of the present embodiment, the number of vertices has been increased by providing means for highlighting the corresponding vertices on the stereo image display and the three-dimensional object display. Even in this case, the relationship between the corresponding points between the stereo image display and the three-dimensional display is easy to understand.
[0033]
[Second Embodiment]
Next, the three-dimensional shape input device of the second embodiment will be described. In the three-dimensional shape input apparatus according to the second embodiment, a vertex position editing function is added to the designated point highlighting function of the first embodiment. That is, by selecting one of the vertices on any of the display images 3, 4, and 5 in FIG. 1 and dragging it with the mouse, the three-dimensional shape can be confirmed while interactively confirming the display in three ways. It can be edited.
[0034]
When the triangular net generation method is taken into consideration, there are two methods for editing the vertex position. That is, a case where a vertex before movement is used for generating a triangular network and a case where a vertex after movement is used. In particular, when the Delaunay network is used for generating the triangular network, the three-dimensional shape generated by the vertex movement operation differs. This will be described later.
[0035]
Hereinafter, the three-dimensional shape input process of the second embodiment will be described. The difference from the first embodiment will be mainly described. FIG. 6 is an explanatory diagram showing the data structure of the second embodiment. In the list element of the point list in FIG. 5A, an item for holding the vertex before the movement is added when the vertex is moved by the editing operation. The surface list in FIG. 5B is the same as that in the first embodiment.
[0036]
The flowchart showing the three-dimensional input processing routine of the first embodiment shown in FIGS. 4 and 5 shifts to the edit (vertex movement) mode and the mode determination in steps S409 to S411 in the edit (vertex movement) mode. By adding steps, the three-dimensional input processing routine of the second embodiment can be configured. If the mode selection is edit mode, edit mode processing is executed.
[0037]
In the second embodiment, in the triangular patch generation (surface list) processing in step S418, if a point before movement exists in the list element of the point list, that is, if the vertex is edited, click here The coordinates of the points are used. However, the coordinates before movement are used only for the connection relationship between the vertices in triangle patch generation, and the triangle patch display after obtaining the connection relationship of the triangle network, the vertex coordinates after movement in the 3D shape generation / display Is used.
[0038]
FIG. 7 is a flowchart showing the edit mode processing routine of the second embodiment. It is assumed that a three-dimensional shape has already been generated to some extent. First, the vertex to be edited is selected (steps S602 and S603).
[0039]
Here, let the selected vertex be Ps. In step S604, information before editing of the vertex to be edited is stored as vertex coordinates before movement. Specifically, one new point list element is generated, and this is assumed to be Ps ′ (in the case of S = 2 in FIG. 6). The contents of the list element of Ps of the point list are copied to this Ps ′, and a pointer to Ps ′ is set to “point before movement”. The contents of Ps ′ are changed in the subsequent processing.
[0040]
Next, when the mouse is moved (dragged) with the mouse button held down, the designated vertices on all windows (3, 4, and 5 in FIG. 1) move with the movement of the mouse. In step S607, the list element of Ps corresponding to the window in which the mouse button is pressed is updated using the mouse coordinates acquired in step S606. For example, if vertex position editing is being performed on the right image, the right coordinate of Ps is updated. In step S608, the three-dimensional coordinates of Ps are recalculated, the three-dimensional coordinates of the list element of Ps are updated, and all display windows are updated. Since the operations in steps S605 to S609 are executed until the mouse button is released, other displays are updated as the point moves on a certain window.
[0041]
Here, when the vertex movement is executed on the three-dimensional display window 5, some special operations are required. That is, if the three-dimensional coordinates can be changed directly on the three-dimensional display window 5 by using a special device such as a three-dimensional mouse, the vertex can be edited. Even in this case, since it is theoretically possible to calculate the two-dimensional coordinates on the left and right stereo images from the three-dimensional coordinates, the vertex position on the stereo image also moves as the vertex position of the three-dimensional display moves. become.
[0042]
When the mouse button is released (step S605), the vertex movement process is terminated when the vertex before movement is used for generating the triangle network (steps S610 and S614). On the other hand, when the moved vertex is used for generating the triangle network, the point list element is updated using the moved vertex coordinates. That is, Ps in the point list is replaced with Ps ′, and the list element of Ps is deleted (step S611). At this point, the content of the pointer to the point before movement is empty. The triangle network is recalculated using the moved vertex (step S612), the display is updated (step S613), and the process ends.
[0043]
It should be noted that the change in the three-dimensional shape associated with the operation differs between the case where the vertices before movement are used for the generation of the triangle network and the case where the vertices after movement are used. When the Delaunay network is used to generate the triangle network, the triangle network is generated so that all the triangles are as close to the regular triangle as possible, that is, no elongated triangle is formed. May change. On the other hand, when using the vertices before moving to generate the triangle network, the vertices before moving are used at the time of generating the triangle network, and the actual coordinates are used after moving, so the connection relationship does not change. However, without being constrained by the Delaunay network, elongated triangles may also be generated. This indicates that it is possible to deal with a concave region that cannot be dealt with by a three-dimensional shape generation method using a simple Delaunay network.
[0044]
As described above, according to the three-dimensional shape input device of the second embodiment, in addition to the vertex emphasis display means of the first embodiment, means for interactively moving the vertex position on each stereo image display And with the movement of vertices on one stereo image display, the corresponding 3D coordinates of the corresponding vertices also have means to perform calculation and 3D display so that there is no contradiction between the mutual coordinates each time they move As a result, it is possible to intuitively generate and edit a three-dimensional shape while confirming a change in the three-dimensional shape of an object with two stereo image displays and three display methods. In addition, the point before editing is used to generate the triangle network by connecting the vertices, and when the actual display and 3D shape data are generated, the edited vertex coordinates are used, thereby creating an unexpected shape associated with the vertex coordinates. Since no change occurred, editing became intuitive and it was possible to deal with concave areas.
[0045]
【The invention's effect】
According to the three-dimensional shape input device of the present invention, the stereo image is displayed by the image display means, the three-dimensional shape data is displayed by the three-dimensional shape data display means, and the left and right images of the stereo image are displayed. The corresponding points are designated by the point designating means, the three-dimensional coordinates are sequentially calculated by the calculating means from the corresponding points and the camera parameters of the stereo image, and the coordinates on the image inputted by the generating means are set in three points at a time. A triangular network is generated by connecting in pairs, and the surface of the object is approximately formed by surface forming means using the generated triangular network and the calculated three-dimensional coordinates, and the generated A triangular mesh is superimposed on the stereo image and the three-dimensional shape data by a superimposing unit, and is displayed on the image display unit and the three-dimensional shape data displaying unit by a vertex specifying unit. It performs the specified vertex, by highlighting means By the vertex designation means According to the specification On the display of the stereo image and the three-dimensional shape data, Since the corresponding vertex is highlighted, Left and right Stereo image When 3D shape It is easy to understand the relations that correspond to each other in three ways of displaying data. For example, By highlighting the vertices corresponding to each other, even if the number of vertices increases, the correlation between the corresponding points between the stereo image display and the three-dimensional shape display is easily understood. In particular, in the case where vertices are complicated, it is difficult to understand which point each vertex corresponds to on these three types of display, and this is effective for the problem that the correspondence between the designated points is likely to be erroneous.
[0046]
According to the three-dimensional shape input device according to claim 2, the vertex position on each stereo image Move Moving means to move, Said Since the calculation means calculates the coordinates of the vertices of the triangle network displayed superimposed on the image display means and the three-dimensional shape data display means while corresponding to each other as the vertices move, in addition to the vertex emphasis display means, Vertex position on stereo image display Move A means for moving and a means for calculating and displaying three-dimensionally so that there is no contradiction between the coordinates of the corresponding vertex each time the vertex moves on one stereo image display. This makes it possible to generate and edit an intuitive and delicate three-dimensional shape while confirming the change in the three-dimensional shape of the object with two stereo image displays and three display methods. In addition, by connecting the vertices, the points before editing are used to generate the triangle network, and when the display and 3D shape data are actually generated, the edited vertex coordinates are used, so that unexpected shape changes accompanying vertex editing can occur. Since it does not occur, editing is intuitive and it is possible to deal with concave areas.
[0047]
According to the three-dimensional shape input method according to claim 3, a stereo image is displayed, three-dimensional shape data is displayed, corresponding points in the left and right images of the stereo image are designated, and the corresponding points and the stereo image are displayed. The three-dimensional coordinates are sequentially calculated from the camera parameters of the image, and the triangle network is generated by connecting the coordinates on the input image in pairs, and the generated triangle network and the calculated 3D coordinates are used to approximately form the surface of the object, and the generated triangular network is superimposed on the stereo image and the 3D shape data, and superimposed on the stereo image and the 3D shape data. Specify the vertices of the triangle network, and according to the specification On the display of the stereo image and the three-dimensional shape data, Since the corresponding vertex is highlighted, Left and right Stereo image When 3D shape It is easy to understand the relations that correspond to each other in three ways of displaying data. For example, By highlighting the vertices corresponding to each other, even if the number of vertices increases, the correlation between the corresponding points between the stereo image display and the three-dimensional shape display is easily understood. In particular, in the case where vertices are complicated, it is difficult to understand which point each vertex corresponds to on these three types of display, and this is effective for the problem that the correspondence between the designated points is likely to be erroneous.
[Brief description of the drawings]
FIG. 1 is an explanatory diagram showing a screen on which a stereo image is displayed.
FIG. 2 is a block diagram showing a configuration of a three-dimensional shape input device.
FIG. 3 is an explanatory diagram showing a data structure used in a three-dimensional shape input processing routine.
FIG. 4 is a flowchart showing a three-dimensional shape input processing routine.
FIG. 5 is a flowchart showing a three-dimensional shape input processing routine continued from FIG. 4;
FIG. 6 is an explanatory diagram showing a data structure of a second embodiment.
FIG. 7 is a flowchart showing an edit mode processing routine of the second embodiment.
[Explanation of symbols]
3, 4, 5 ... Window image
11 ... Input / output section
12 ... Hard disk device
13 ... Main memory
14 ... CPU
15 ... Bitmap display
16 ... Keyboard
17 ... mouse

Claims (3)

Image display means for displaying a stereo image;
3D shape data display means for displaying 3D shape data;
Point designating means for designating corresponding points in the left and right images of the stereo image;
Calculation means for sequentially calculating three-dimensional coordinates from the corresponding points and camera parameters of the stereo image;
Generating means for generating a triangular network by connecting the coordinates on the input image in groups of three points; and
Surface forming means for approximately forming the surface of the object using the generated triangular network and the calculated three-dimensional coordinates;
Superimposing means for superimposing the generated triangular network on the stereo image and the three-dimensional shape data;
Vertex designating means for designating the vertexes of the triangle net superimposed on the image display means and the three-dimensional shape data display means;
A three-dimensional shape input apparatus comprising highlighting means for highlighting vertices corresponding to each other on display of the stereo image and the three-dimensional shape data in accordance with designation by the vertex designation means . Moving means for moving the vertex position on each stereo image;
2. The calculation unit according to claim 1, wherein the calculation unit calculates the coordinates of the vertices of the triangle network superimposed on the image display unit and the three-dimensional shape data display unit while corresponding to each other as the vertices move. 3D shape input device. Display a stereo image,
Display 3D shape data,
Specify corresponding points in the left and right images of the stereo image,
3D coordinates are sequentially calculated from the corresponding points and the camera parameters of the stereo image,
Create a triangle network by connecting the coordinates on the input image in pairs,
Using the generated triangular network and the calculated three-dimensional coordinates, approximately form the surface of the object,
Superimposing the generated triangular network on the stereo image and the three-dimensional shape data;
Specify the vertices of the triangle network superimposed on the stereo image and the three-dimensional shape data,
A three-dimensional shape input method characterized by highlighting mutually corresponding vertices on the display of the stereo image and the three-dimensional shape data in accordance with the designation.

Images